Sagnyx

One step ahead of ordinary.
AI / Tech / Workflows

[Sagnyx @sagnyx]

AN 18-YEAR-OLD IN SHENZHEN BUILT A ROBOT THAT USES A FLY'S BRAIN TO COPY HIS HANDWRITING

The entire machine costs about 400 yuan in Taobao parts.

Two stepper motors.
A 6-yuan gel pen.
A 3D-printed pen head.
Then he gave it something to learn from: 40 pages of his own notebooks containing roughly 11,000 characters.
Not just the letters.

The loops.
The slant.
The spacing.
Even the tiny inconsistencies that appear when you've been writing for 14 hours.

Then he added three separate "brains."

The first generates the answers using ChatGPT.
The second contains three years of his class notes in a vector store. It takes those answers and rewrites them in his vocabulary, sentence length and even his mistakes - deliberately adding two errors per page.
The third is the strange part.
He took the FlyWire connectome: a complete map of a fruit fly's head containing 139,255 neurons and 54.5 million synapses, divided into 7,062 layers by researchers and released publicly.

He uses that biological circuit as the robot's motor controller.

The same kind of wiring that once helped control a fly's wing beat is now controlling the movement of a pen.

Most handwriting plotters reveal themselves almost immediately.

Give them four seconds and you can see it.

The strokes are too consistent.
The spacing is too perfect.
Everything looks mechanical.

This robot does the opposite.

The fly-based controller introduces a 0.3 mm wobble into the strokes - almost like the movement of a tired human wrist late at night.

And that imperfection is exactly what makes it convincing.

62 assignments later.
4 teachers.
Average grade: 91.

One teacher even wrote in the margin:
“Your handwriting has gotten more natural this term.”
The fly has been dead for years.

Its wing circuitry is still moving a pen at midnight.

[Sagnyx @sagnyx]

The fruit fly may have just given neuromorphic chips something they’ve been missing: a real brain blueprint.

Scientists spent 20 years mapping the entire nervous system of Drosophila melanogaster, down to every single connection.

The result is wild:

166,700 nerve cells
124.2 million synapses
Every connection mapped and published.
And this isn’t just sitting in a database.
Researchers are already running the same connectome on neuromorphic chips like Intel’s Loihi 2.

They’re using it to build working models of how a real brain makes decisions - including something as simple as deciding whether to eat or groom itself.

A fruit fly’s brain could become one of the most useful blueprints for building machines that process information more like a real brain.

[Sagnyx @sagnyx]

A dead fly just got a second body.

Inside a robotic body now sit 166,700 fruit fly neurons connected through more than 25 million biological connections.

And the strange part is that the robot looks almost like the fly those neurons came from.

Researchers gave the neurons cameras, wings and motors - but didn’t program the system with a normal set of flight behaviors.

Instead, they let the biological network figure out what the incoming signals meant.

A light moves → the head turns.

The airflow changes → the wings adjust.

Another fly appears on camera → the robot begins following it.

That’s what makes this so unsettling.

The nervous system doesn’t need a huge amount of hardware before some of the behaviors it evolved for start becoming useful again.

Nature spent hundreds of millions of years developing this tiny biological system to react to movement, maintain stability and chase what moves.

Now that same wiring has been connected to a body made of metal.

The original fly is gone.

But its nervous system is still trying to fly.

[Sagnyx @sagnyx]

A Stanford student gave a robot 166,700 fruit fly neurons - and it now does homework for 30 different students.

None of them have needed to pick up a pen.

The machine is a black aluminum frame holding a blue ballpoint pen. On the page, it writes fractions, square roots and entire lines of math in handwriting that looks surprisingly human. In the corner of the screen, the fly’s neurons are glowing.

Before writing, each student gives the system a sample of their handwriting.

The fly’s neural network learns the tiny differences: the angle of the letters, the spacing between words, the size of loops, even the way someone’s handwriting gets lazier near the end of a line.

Then it reproduces that student’s hand.

30 notebooks.
30 unique handwriting styles.
166,700 neurons doing the work.
The entire fly brain is smaller than a poppy seed.
A human brain has more than 500,000 times as many neurons.

Yet this tiny biological system doesn’t get tired, doesn’t smear the ink, and never confuses which student puts a line through their 7s.

One teacher checks 30 notebooks.

A fruit fly wrote all of them.

[Sagnyx @sagnyx]

Google and HHMI Janelia spent a decade mapping an entire fruit fly nervous system.

Then they released the whole thing for anyone to use.

166,700 neurons.
24.5 million synapses.
Connections mapped from the eyes all the way to the legs.
One student took that dataset and did something unexpected: she rebuilt the wiring as a leaky integrate-and-fire neural simulation and turned it into an AI agent.
A camera watches her hand.

Her finger movements are translated into signals for the fly's simulated sensory neurons. The resulting motor neuron activity determines what a robot does on the other side of the screen.

▪ Pinch → running implementation
▪ Point → new implementation
��� Pull back → heavy lifting

And the progression gets even stranger.

Just 12 days earlier, this was nothing more than a dataset.
Then the simulated brain played DOOM.
Then Chrome Dino.

Then it controlled a walking Strandbeest.

Now it's being used as an agent by a student.

It's not a living fly.

It's a computer simulation built from the actual wiring of a real fly's nervous system - with the connection between her hand and those neurons built by her.

The fly died years ago.

Its wiring is still taking orders.

[Sagnyx @sagnyx]

@lynk0x If you don't want to lose money, it's better not to keep track of it at all😁

[Sagnyx @sagnyx]

$244K from stolen Roblox accounts at just 18.

Cyber police from Ukraine’s National Police paid a visit to a young guy from Drohobych who had assembled a team and pulled off a crazy scheme targeting Roblox players.

The guy took advantage of kids’ trust. He uploaded software disguised as cheats to various gaming forums, promising players an advantage in Roblox.

On those same forums, he found accomplices. The software wasn’t just a cheat - it contained a stealer. Kids who downloaded it hoping to get an unfair advantage ended up losing their Roblox accounts instead.

He picked the perfect target audience: Roblox is mainly played by kids and teenagers, many of whom are trusting and actively looking for cheats and free stuff.

The stolen accounts were then processed through a program they had developed. It checked whether the account credentials still worked and immediately estimated the account’s “value” based on things like rare skins, premium currency, and other valuable items.

They managed to steal more than 610,000 accounts. The most valuable ones were selected and sold on Russian marketplaces.

In total, they made more than $244,000.

During the searches, police seized computers, flash drives, mobile phones, €2,500 and $35,000 in cash.

[Sagnyx @sagnyx]

AI OR REAL GTA VI?👀

GTA 6 is everywhere right now, and every new video makes you wonder - are we looking at actual gameplay or another incredibly realistic AI generation?

This video has pretty much everything: a packed Vice City beach, nighttime police scenes, a ride through the swamps on an airboat, cars, a plane flight, city streets, a nightclub, a store and even a bank robbery.

And the craziest part is that it all looks like one continuous gameplay session, rather than just a bunch of separate AI-generated clips.

But is this actually GTA VI?

I’m not 100% sure, so let’s figure it out together. What do you think - REAL or AI? 👀

#GTA6 #GTAVI

[Sagnyx @sagnyx]

@juKozimq In general yes, would like better

[Sagnyx @sagnyx]

@_GTAVI_ the picture looks average so far

[Sagnyx @sagnyx]

@_GTAVI_ This is already a ready-made mission in GTA 6: run from Rockstar and the FBI while you're playing through the story yourself

[Sagnyx @sagnyx]

@StarPlatinum_ I’m laughing even more at the fact that people are comparing this to $TRUMP and $MELANIA - but the only difference is that nobody knew about the launch of those tokens except the insiders.

Here, literally everyone is already waiting for the token

[Sagnyx @sagnyx]

@jonny_pixels_ As soon as I hear this music and see this train station, I’m instantly transported back to 2010… Mafia 2 forever in my heart

[Sagnyx @sagnyx]

GTA 6 is starting to look less like a game and more like a movie you can actually play.

The lighting. The water. The animations. The insane level of detail even on a random street.

If the actual game looks even remotely this good, I’ll probably spend hours just driving around Vice City, playing good music and enjoying the vibe. 😂

[Sagnyx @sagnyx]

@prokopkor Damn, this ain’t a game anymore - this is pure vibe 🔥

[Sagnyx @sagnyx]

@HeyRohhit Sounds like a must-watch. Gonna go check it out

[MegaETH @megaeth]

The First Real-Time Blockchain.

Built for you. Priced by you.

Our public sale on Sonar by Echo, starting at $1M FDV, makes you the largest stakeholder in our network. [thread]

[OKX Wallet @wallet]

DOOD is launching soon⌛

We've partnered with @doodles to give our users exclusive early access to 100,000,000 $DOOD 🤝

Keep reading for the alpha👇

[doodles @doodles]

$DOOD will be launching on Solana 🔜

we’ve optimized for a smooth experience and will be airdropping $DOOD to eligible communities, whether you’re a current Doodles NFT holder or a New Blood community member.

here's how to get your airdrop ↓

[Layer3 @layer3xyz]

v2 is here –– it's our largest update yet.

Here's the alpha 👇